Evidence based medicine-workbook


Published on

Published in: Health & Medicine, Education
  • Be the first to comment

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

Evidence based medicine-workbook

  1. 1. Evidence-based Medicine Workbook Finding and applying the best research evidence to improve patient care ?Written by Paul Glasziou, Chris Del Mar and Janet Salisbury
  2. 2. Evidence-based Medicine WorkbookFinding and applying the best evidence to improve patient care Based on workshops presented by: Professor Paul Glasziou Professor Chris Del Mar Centre for Centre for Evidence-Based Practice General Practice The University of Queensland, The University of Queensland, Australia and Oxford, UK Australia Concept development and writing by Janet Salisbury, Biotext, Canberra, Australia
  3. 3. © BMJ Publishing Group BMJ Books is an imprint of the BMJ Publishing GroupAll rights reserved. No part of this publication may be reproduced, stored ina retrieval system, or transmitted, in any form or by any means, electronic,mechanical, photocopying, recording and/or otherwise, without the priorwritten permission of the publishers.First published in by BMJ Books, BMA House, Tavistock SquareLondon WCH JRwww.bmjbooks.comBritish Library Cataloguing in Publication DataA catalogue record for this book is available from the British LibraryISBN    This workbook has been developed from workshop handouts used at evidence-based medicine workshops run by the Centre for Evidence-Based Practice,Centre for General Practice, The University of Queensland. Its developmentwas supported by the Australian Department of Health and Ageing under thePrimary Health Care Research Evaluation and Development Strategy.Production management by Biotext, CanberraDesign by Peter Nolan, Clarus Design, CanberraCartoons by Ian Sharpe, CanberraPrinted and bound in Spain by Graphycems, Navarra
  4. 4. Introduction to this workbookMedical practitioners, particularly GPs, are overloaded with information.They simply cannot keep up with reading all the scientific literature and otherinformation that arrives on their desk every week. Even when they have time toread some of it, it is difficult to identify which information will be most usefulin clinical practice and to recall the most up-to-date findings when they needthem.But every day doctors encounter questions that need to be answered in orderto make the best decisions about patient care. This is where ‘evidence-basedmedicine’ (EBM) comes in. The aim of this workbook is to introduce generalpractitioners and other health care professionals to the concept of EBM and toshow them simple methods to find and use the best evidence to answer theirclinical questions.The workbook is practical and interactive, and will develop your skills in:• asking clinical questions• searching for answers• using the answers to make clinical decisions.At the end of this workbook, we hope that you will feel confident that you canfind the best quality evidence for almost any clinical question that comes yourway and, with a little practice, use it to improve your clinical practice, all withina few minutes. iii
  5. 5. How to use this workbookThis workbook has been based on the evidence-based medicine workshopsrun by the Centre for Evidence-Based Practice and Centre for General Practice,University of Queensland and contains information and exercises to help youlearn how to use evidence-based medicine (EBM) in your clinical practice.The workbook is divided into three main parts:Part 1 (purple) contains an introduction to EBM and some clinical examples to show how it can be applied.Part 2 (blue) describes the practical application of EBM. It is subdivided into five modules, each describing an important stage in the EBM process (how to formulate a question, how to track down the best evidence, how to critically appraise the evidence, how to apply the evidence and how to evaluate your progress).Part 3 (yellow) contains information on useful internet sites and EBM resources and a number of useful articles for further reading.If you attend one of our workshops, you will find that this workbook containsall the information that will be presented during the workshop. This means thatyou do not need to worry about writing down a lot of notes or copying downslides. Just relax and concentrate on the sessions. There are spaces in the kit foryou to write down information during the interactive sessions and record theresults of your EBM activities during the day.The workbook has also been designed as a plain English resource document foranyone who is interested in learning more about EBM to study at their leisureor share with colleagues in small group training sessions.In either case, we hope that you find it useful.So that we can improve the workbook in future editions, please provide us withyour feedback.iv
  6. 6. ContentsPart 1: Introduction to evidence-based medicine 1 What is evidence-based medicine? 3 Some evidence-based cases 14Part 2: The steps in EBM 21 EBM step 1: Formulate an answerable question 23 EBM step 2: Track down the best evidence 43 EBM step 3: Rapid critical appraisal of controlled trials 61 EBM step 4: Apply the evidence 87 EBM step 5: Evaluate the effectiveness and efficiency of the process 93Part 3: Resources and further reading 97 Useful sources of evidence 99 Further reading 103 Index 130 v
  7. 7. What is evidence-based medicine? “ … the integration of best research evidenceClinical practice is about making choices. Shall I order a test? Should I treat the with clinical expertise andpatient? What should I treat them with? The decision depends on the doctor’s patient values”knowledge, skills and attitudes, and on what resources and tests are available.The patient’s concerns, expectations and values also need to be taken into – Dave Sackettaccount.The term ‘evidence-based medicine’ (EBM) was first used by a Canadian,David Sackett and his collegues at McMaster University in Ontario, Canadain the early 1990s. They have subsequently refined the definition of EBM asintegrating the best research evidence with clinical expertise and patient valuesto achieve the best possible patient management. EBM is about trying toimprove the quality of the information on which decisions are based. It helpspractitioners to avoid ‘information overload’ but, at the same time, to find andapply the most useful information. of clinical decisio Sackett DL, Stranss SE, Richardson WS et al. ents nm Evidence-Based Medicine. How to practice on and reach EBM. Edinburgh: Churchill Livingstone, 2000. p ak Photograph reproduced with permission. m ing Co Steps in EBM Doctors Patient knowledge Clinical decisions EBM uses a series of steps: values, of evidence, concerns, skills, expectations 1. Formulate an answerable attitude question. 2. Track down the best Health system evidence of outcomes access rules Concern about available. (PBS, Medicare litigation funding, etc) 3. Critically appraise the evidence (ie find out how good it is). 4. Apply the evidence (integrate the results withEBM, which has largely replaced the older term ‘clinical epidemiology’, is clinical expertise andsometimes also called ‘evidence-based practice’. This latter term highlights patient values).the important point that the ‘evidence’ that we are talking about is empiricalevidence about what actually works or doesn’t work in practice. It is not 5. Evaluate the effectivenessscientific evidence for a mechanism of action (such as a biochemical pathway, and efficiency of thephysiological effect or anatomical feature). Many factors affect the outcomes of process (to improve next time).medical activities: the underlying mechanism is only one of them. EBM is aboutactual clinical outcomes. 3
  8. 8. Why do we need EBM?Unfortunately, there is a large information gap between research and clinicalpractice. Because so much research is published all the time, cliniciansunderstandably are unaware of most of it, or do not have the ‘tools’ to assessits quality. Researchers, on the other hand, do not understand the informationneeds of clinicians and continue to present their work in a way that is noteasily accessible to busy practitioners. In 1972, British epidemiologist ArchieCochrane highlighted the fact that most treatment-related decisions werebased on an ad hoc selection of information from the vast and variable qualityscientific literature, on expert opinion, or, worse of all, on trial and error. Professor Archie Cochrane was a medical researcher in the UK who contributed to the development of epidemiology as a science. In an influential book published in 1972 (Effectiveness and Efficiency), he drew attention to the great collective ignorance at that time about the effects of health care. He recognised that doctors did not have ready access to reliable reviews of available evidence. In a 1979 article he said: ‘It is surely a great criticism of our profession that we have not organised a critical The Cochrane Collaboration summary, by speciality or subspeciality, adapted periodically, of all relevant was found in response to randomised controlled trials.’ Cochrane’s call for systematic, up-to-date reviews of References: all relevant randomised controlled trials of health care. Cochrane AL (1972). Effectiveness and Efficiency. Random Reflections on Health Services, In the early 1990s, funds were Nuffield Provincial Hospital Trust, London (reprinted in 1989 in association with the provided by the UK National British Medical Journal). Health Service to establish a Cochrane AL (1979). 1931–1971: A critical review, with particular reference to the medical Cochrane Centre in Oxford. profession. In: Medicines for the Year 2000, Office of Health Economics, London. The approach was further outlined at an international meeting organised by the New York Academy of Sciences in 1993 and at the first CochraneCochrane proposed that researchers and practitioners should collaborate Colloquium in Octoberinternationally to systematically review all the best clinical trials (that is, 1993, when ‘The Cochranerandomised controlled trials, or RCTs), specialty by specialty. His ideas were Collaboration’ was founded.gradually taken up during the 1980s by Iain Chalmers and one of the first areasof clinical practice to be reviewed in this way was care during pregnancy and http://www.cochrane.orgchildbirth. Systematic reviews of RCTs of different aspects of obstetric care soon Cochrane logo produced withshowed some anomalies between the clinical trial evidence and established permission from The Cochranepractice. This highlighted the gaps that existed between research and clinical Collaborationpractice and started to convince some doctors of the benefits of an evidence-based approach to bridge this gap.4
  9. 9. This work has been continued though the international ‘CochraneCollaboration’, which publishes systematic reviews of RCTs electronically inthe Cochrane Library of Systematic Reviews. This database, which we willbe looking at in detail later in the workshop, is available free online in manycountries:http://www.cochrane.org and follow the prompts. CORTICOSTEROIDS FOR PRETERM BIRTH 1972 A RCT was reported showing improved outcomes for preterm babies when mothers were given a short course of corticosteroid before the birth. 1972–89 Six more RCTs were published which all confirmed 1972 findings. During this time, most obstetricians were still unaware that corticosteroid treatment was so effective and so did not treat women about to give birth early with corticosteroids. 1989 First systematic review published. 1989–91 Seven more studies reported. Corticosteroid treatment reduces the odds of babies dying from complications of immaturity by 30 to 50% but thousands of babies have died or suffered unnecessarily since 1972 because doctors did not know about the effectiveness of the treatment. 5
  10. 10. The flecainide storyThe history of the use of the drug flecainide to treat heart attacks in the UnitedStates in the 1980s is a dramatic example of the gap between research andclinical practice, and the reliance on evidence of a mechanism rather than anoutcome. In 1979, the inventor of the defibrillator, Bernard Lown, pointed outin an address to the American College of Cardiology that one of the biggestcauses of death was heart attack, particularly among young and middle-agedmen (20–64-year-olds). People had a heart attack, developed arrhythmiaand died from the arrhythmia. He suggested that a ‘safe and long-actingantiarrhythmic drug that protects against ventricular fibrillation’ would savemillions of lives.In response to this challenge, a paper was published in the New England Journalof Medicine introducing a new drug called flecainide — a local anaestheticderivative that suppresses arrhythmia. The paper described a study in whichpatients who had just had heart attacks randomly received placebo orflecainide and were then switched from one to the other (a cross-over trial).The researchers counted the number of preventricular contractions (PVCs) asa measure of arrhythmias. The patients on flecainide had fewer PVCs than thepatients on placebo. When the flecainide patients were ‘crossed over’ to theplacebo, the PVCs increased again. Suppression of arrythmias in 9 patients (PVCs = preventricular contractions) 50 45 40 35PCVs/12 hours 30 25 20 15 10 5 0 Placebo Flecainide Placebo FlecainideThe conclusion was straightforward: flecainide reduces arrythmias andarrythmias cause heart attacks (the mechanism); therefore, people who havehad heart attacks should be given flecainide. The results were published in theNew England Journal of Medicine and flecainide was approved by the UnitedStates Food and Drug Adminstration and became fairly standard treatmentfor heart attack in the United States (although it did not catch on in Europe orAustralia).6
  11. 11. Almost immediately after the first trials were complete, however, otherresearchers had started gathering information on the survival of the patients Overall, the flecainide story(the outcome) instead of the PVC rate (the mechanism). This showed that over raises two important issues:the 18 months following treatment, more than 10% of people who were givenflecainide died, which was double the rate of deaths among a placebo group. • We need a better wayIn other words, despite a perfectly good mechanism for the usefulness of to find information,flecainide (it reduces arrhythmias), the drug was clearly toxic and, overall, did even when we do notmuch more harm than good. know that we need it. In other words, up- Cardiac arrythmia suppression trial (CAST) to-date, good-quality research findings need 100 to be available to all medical practitioners on a routine basis. 95 control • The type of research is important. We must % Alive 90 move away from a flecaininde traditional mechanistic approach and look 85 for empirical evidence of effectiveness using 80 a clinically relevant 0 200 400 600 outcome (eg survival, improved quality of life). DaysUnfortunately, because the initial studies had been widely published in medicaltexts, it was a long time before doctors caught up with the subsequent pooroutcome data, which did not attract as much attention. Meanwhile, about200,000 people were being treated with flecainide in the United States by1989. Based on the trial evidence, this would have caused tens of thousandsof additional heart attack deaths due to the use of flecainide. Although there References (flecainide):was published information, doctors were systematically killing people with Anderson JL, Stewart JR, Perry BA etflecainide because they did not know about the good quality outcome-based al (1981). Oral flecainide acetateresearch. for the treatment of ventricular arrhythmias. New England JournalWhat does the flecainide example tell us? of Medicine 305:473–477. Echt DS, Liebson PR, Mitchell LB et alIn the flecainide example, the initial research was widely disseminated because (1991). Mortality and morbidityit was based on a traditional mechanistic approach to medicine and because in patients receiving ecainide,it offered a ‘cure’. The subsequent outcomes research may not have been flecainide, or placebo. The Cardiacwidely disseminated because it was counterintuitive and negative in terms Arrythmia Suppression Trial. Newof a potential treatment. Doctors continued to prescribe flecainide because England Journal of Medicine 324:they believed that it worked. They did not know that they needed to look for 781–788.further information. Moore TJ (1995). Deadly Medicine, Simon and Schuster, New York. 7
  12. 12. So much evidence, so little time ‘Kill as few patients asDoctors need to be linked to the medical research literature in a way that possible’allows them to routinely obtain up-to-date, outcomes-based information.However, most medical practitioners, particularly GPs, are overloaded with A book by physician andinformation. Unsolicited information received though the mail alone can medical humorist Oscaramount to kilograms per month and most of it ends up in the bin. London called ‘Kill as Few Patients as Possible’ gives a setThe total number of RCTs published has increased exponentially since the of ‘rules’ for clinical practice.1940s. A total of 20,000 trials are published each year (with over 300,000 intotal) and approximately 50 new trials are published every day. Therefore, to Rule 31 offers some advice onkeep up to date with RCTs alone, a GP would have to read one study report how to keep up to date with medical research:every half hour, day and night. In addition to RCTs, about 1000 papers are alsoindexed daily on MEDLINE from a total of about 5000 journal articles published ‘Review the world literatureeach day. fortnightly’. The amount of medical research 5000 per day 2000000 1800000Medical articles year 1600000 1400000 1200000 1000000 800000 1000 per day 600000 400000 50 per day 200000 0 Trials MEDLINE TOTAL 25000 20000 Number of RCTs year 15000 10000 5000 Parliament house flagpole (81 m) 0 A year of 50 55 60 65 70 75 80 85 90 95 MEDLINE indexed journals Year8
  13. 13. At best, most GPs give a selective sample of the literature a cursory review,but very little is properly assessed and almost none influences what they do inpractice.Doctors may feel guilty, anxious or inadequate because of this (see the JASPAcriteria), but it is not their fault — there is just too much of it. There needs tobe a better way. JASPA criteria (journal associated score of personal angst) Can you answer these five simple questions: J: Are you ambivalent about renewing your journal subscriptions? A: Do you feel anger towards particular authors? S: Do you use journals to help you sleep? P: Are you surrounded by piles of periodicals? A: Do you feel anxious when another one comes through the letterbox? Score (Yes = 1; No =0): 0 anyone who scores zero is probably a liar! 1–3 normal range >3 sick, at risk for ‘polythenia gravis’ and related conditions Modified from: Polythenia gravis: the downside of evidence-based medicine. British Medical Journal (1995) 311:1666–1668. 9
  14. 14. How do doctors try to overcome information overload?Write down some education activities that you and your organisation engagein and how much time you spend on them. Your education activities How much time do you spend on each?You have probably included a selection of activities including attendinglectures and conferences, reading journals, textbooks and clinical practiceguidelines, electronic searching, clinical attachments and small group learning.You may also have included talking to colleagues or specialists. But everyonehas the same problem of keeping up to date and your colleagues may be outof date or just plain wrong. If they have got the information from somewhereelse, you need to know where they got it so that you can check how good it is.Textbooks are always about 5–10 years out of date.Faced with all the alternatives, how do you actually choose what to do in yourcontinuing education time? If you are honest, your choice probably dependson what you are already most interested in rather than what you don’t knowabout.10
  15. 15. Continuing medical education (CME) has been a mainstay of doctors’professional development but no-one has ever shown that it works. When Conclusions ofdoctors choose their courses, they choose things that they think they need to CME trialknow about. But as we have seen, the most important information is what theydon’t know they need! We need a system to tell us we need to know something. 1. When you want CME, you don’t need it.In a trial of CME, a random sample of GPs were asked to rank 18 selected 2. CME only works when youconditions into either a ‘high preference’ set, for which they wanted to receive don’t want it.CME or a ‘low preference’ set for which they did not want further education. 3. CME does not causePhysicians with similar rankings were paired and randomised to either: general improvements in the quality of care.• a control group whose CME was postponed for 18 months; or• an experimental group who received CME at once for their high preference topics and were also provided with training materials for their low preference topics and asked to promise to study them.The outcomes were measured in terms of the quality of clinical care (QOC)provided by each of the physicians before and after CME (determinedfrom clinical records). The results showed that although the knowledgeof experimental physicians rose after their CME, the effects on QOC weredisappointing with a similar (small) increase in QOC for both the experimentaland control groups for their high preference conditions.By contrast, for low preference conditions, QOC rose significantly for the Reference:experimental physicians but fell for the control group. Sibley JC, Sackett DL, Neufeld V et al (). A randomised trialA review of didactic CME by Davies et al (1999) also concluded that formal of continuing medical education.sessions are not effective in changing physician performance (see Part 3: New England Journal of MedicineResources and further reading). :–. 11
  16. 16. Overall, as we have seen, there is too much information but we still needit. The quality of most of the information is also very poor: most published Doctors’ informationinformation is irrelevant and/or the methods are not good. Finding the high- needsquality evidence is like trying to sip pure water from a water hose pumpingdirty water, or looking for ‘rare pearls’. Study 1 (interns) • 64 residents in 2 hospitals High quality/relevant data — pearls were interviewed after 401 consultations High • They asked an average of 280 questions (2 per High quality, 3 patients) relevant (=`pearls) • Pursued an answer for 80 Clinical relevance questions (29%) • Others not pursued because of: – lack of time, or – because they forgot the question • Souces of answers: Low – textbooks (31%) – articles (21%) Low High – consultants (17%) Validity Study 2 (GPs)How many questions can doctors answer each day? • 103 GPs in Iowa collected questions over 2.5 daysMany questions arise every day as a result of seeing people in clinical practice. • A total of 1101 questionsTwo papers have been published about this: one of interns in a hospital setting were collectedand one of GPs. In both cases, the researchers asked the doctors to note every • Pursued answers fortime a question arose and what information they needed. 702 questions (64%) • Spent less than 2 minutesThe study of 100 GPs showed that they each wrote down about 10 questions pursuing an answer usingover a two and a half day period. The GPs tried to find answers for about half readily available print andof these. The most critical factor influencing which questions they followed human resourcesup was how long they thought it would take to get an answer. If the doctor • Only 2 questions (0.2%)thought the answer would be available in less than a couple of minutes, they led to a formal literaturewere prepared to look for it. If they thought it would take longer, they would searchnot bother. Only two questions in the whole study (ie 2/1000) were followedup using a proper electronic search. References: Green ML, Ciampi MA and Ellis PJ (2000). Residents’ medical information needs in clinic: are they being met? American Journal of Medicine 109: 218–233. Ely JW, Osheroff JA, Ebell MH et al (1999). Analysis of questions asked by family doctors regarding patient care. British Medical Journal 319:358–361.12
  17. 17. Information gathering Balance yourThere are two ways in which we all get information: ‘push’ and ‘pull’. information:‘Push’ new relevant and valid results ‘pull’ and ‘push’The ‘push’ method of getting our information is when we extract informationfrom the variety of sources that we receive, across a wide spectrum of topicsthat may interest us. This is sometimes called ‘just in case’ learning.For EBM, the best sources for the ‘push’ approach to improving knowledge arewhere the ‘pearls’ have already been selected from the rest of the lower qualityliterature. Some good sources of information where this has been done include:Evidence-Based Medicine — a journal containing information from cliniciansaround the world who spot articles that pass rigorous validity criteria and areimportant to clinical practice. The journal is published every two months andhas no original articles but it gives a condensed version of the original paper. ‘Push’ is when we receive information from a varietyAlso available on the internet at: http://www.evidence-basedmedicine.com of sources and on a variety of topics and extract whatClinical Evidence — is a compendium of evidence based literature searches. It we think we need foris updated and published every six months as a book and CD. It is arranged by our practice (‘just in case’specialty and just states the best existing evidence for an intervention. If there learning).is no evidence, it says so. It does not include opinions or consensus guidelines.The editors decide what questions are relevant but the book is based on whatdoctors need. Doctors can look up information when they need it (the ‘pull’method of obtaining information).Clinical Evidence is available on the internet at:http://www.clinicalevidence.com‘Pull’ answers in less then two minutesIn this workbook we focus on how to formulate questions and ‘pull’ answersout of the literature in less than two minutes! This is sometimes called ‘just in ‘Pull’ is when wetime’ learning. In the next sections we will look at some case studies where deliberately seekEBM methods were used and then find out how to frame a question to make information to answer ait easier to answer. Then we will learn about how to use MEDLINE and the specific question (‘just inCochrane databases to electronically search for the information we need and, time’ learning).finally, how to use the results. 13
  18. 18. Some evidence-based casesIn this session, we will discuss several case studies that show how EBM canhelp in a range of clinical situations. You can then think of your own clinicalquestion which you would like to answer at the workshop.Case study 1: persistent coughA 58-year-old who was visiting her GP about another matter, said, as an aside:‘Can you do anything about a cough?’ She had had a persistent cough for 20 yearswith various treatments but no cure. She had been referred twice to physicians.The GP searched PubMed (the web-based version of MEDLINE) using ‘ClinicalQueries’, which is a category of PubMed designed for clinicians (see pages52–54). The search for persistent cough revealed that the most common causesof a persistent cough are:• postnasal drip• asthma• chronic bronchitisThe GP thought the cough was most likely to be due to asthma, and prescribedappropriate treatment for asthma as a first line of treatment. The patientthought she had already tried that treatment and that it did not work buttried it again anyway, without success. However, the search also showed thatgastro-oesophageal reflux is a less common but possible cause of persistentcough (10% of cases), which the GP had not known before. The GP thereforerecommended the patient to take antacids at night and raise the head of herbed. After one week her cough disappeared for the first time in 20 years andhas not come back since.How did EBM help?This case raises interesting questions of what doctors ‘should’ know. It waswritten up in the British Medical Journal and published as an example of howEBM can help GPs. However, some physicians wrote in saying that ‘everyoneshould know’ that gastro-oesophageal reflux was a possible cause of cough.The author replied that although respiratory physicians might know thisinformation, GPs did not necessarily know it. An anaesthetist wrote in to saythat after reading the article he had been treated for gastro-oesophageal reflux, Reference:which had cured a cough he had had for 30 years! Glasziou P (1998). Evidence based case report: Twenty year coughConclusion: EBM can help you find the information you need, whether or not in a non-smoker. British Medicalyou ‘should’ already know it. Journal 316:1660–1661.14
  19. 19. Case study 2: dog bite Empirical measures ofA person came to the clinic with a fresh dog bite. It looked clean and the outcomesGP and patient wondered whether it was necessary to give prophylacticantibiotics. She searched MEDLINE and found a meta-analysis indicating that Outcomes are commonlythe average infection rate for dog bites was 14% and that antibiotics halved this measured as absolute riskrisk. In other words: reduction (ARR), relative risks (RR) and number needed to• for every 100 people with dog bites, treatment with antibiotics will save 7 treat (NNT). from getting infected; or The risk of infection after dog• treating 14 people with dog bites will prevent one infection. bite with no antibioticsThe second number (14) is called the ‘number needed to treat’ (NNT). = 14% (0.14)The GP explained these figures to the patient, along with the possible The risk of infection after dogconsequences of an infection, and the patient decided not to take antibiotics. bite with antibioticsOn follow-up it was found that the patient did not get infected. = 7% (0.07) The ARR for antibioticHow did EBM help? treatmentIn this case EBM helped because the empirical data were easy for the patient = 14 – 7 = 7%to understand and she could participate in the clinical decision. As the culture (That is, 7 people in every 100of health care changes further towards consumer participation in health care treated will be saved fromdecision making, patients will demand this type of information. infection.) NNT = 100/7Reference: = 14Cummings P (1994). Antibiotics to prevent infection in patients with dog bite wounds: a meta- (That is, you would need to analysis of randomized trials. Annals of Emergency Medicine 23:535–540. treat 14 dog bite patients with antibiotics to prevent 1 infection.) RR of infection with antibiotics compared to without antibiotics = 0.07/0.14 = 0.5 (50%) NOTE : It is best to quote the ARR or NNT in discussions with patients. RR is harder to put into context because it is independent of the frequency of the problem (the ‘event rate’), in this case, the rate at which people with dog bites get infected. Further information on these measures is given in EBM Step 4 (Rapid critical appraisal). 15
  20. 20. Case study 3: microscopic blood in the urine Study 1One of us, then a healthy 47-year-old male, was acting as a patient in amedical exam. The students accurately found microscopic traces of blood in 10,000 men were screened.his urine when they tested it. He went to his GP and was retested a month About 250 (2.5%) hadlater. The blood was still there. The GP suggested conventional investigation: haematuria. These men werean ultrasound and cystoscopy. It was time for a search of the literature for asked to visit their GP andevidence of the effectiveness of these procedures. about 150 (60%) did so. Of those, only three had a serious problem. Of these:He searched for a cohort study of 40–50-year-olds with haematuria with long-term follow-up and for RCTs of screening for haematuria. He used the search • 2 had bladder cancercategories ‘prognosis’ and ‘specificity’ and the search terms ‘haematuria OR • 1 had reflux nephropathy.hematuria’. He got 300 hits. Two papers were very relevant (see box). This shows that there is aboutThe presenter concluded that blood in urine is not a good indicator of bladder a 1 in 50 chance of having acancer and did not have the cystoscopy test. serious disease.How did EBM help? Study 2The lesson from this case concerns the practical versus the empirical. Doctorstend to think along the lines of: A urine test to 20,000 men as part of a work-based personal Blood does not belong in the urine so it must be coming from somewhere. It health appraisal. Follow-up could be coming from a potentially serious cause, such as bladder cancer. studies of the men who were positive for haematuria found three cancers per year, or 1.5Empirical questions, on the other hand, ask about outcomes — in this case cancers per 1000 person-years.whether conventional investigation leads to better health outcomes. In However, the people who werethis case, the evidence (surprisingly) showed no benefit from this, because not found to have haematuriamicroscopic haematuria seems to be no more prevalent among those who were also followed up and thelater develop urological cancer than those who do not. Once again, this allows rate of cancer was exactly thepatients to participate much more fully in clinical decisions. same as for the people with haematuria. EBM can also help reduce litigation This case raises the issue of possible litigation. What if the patient is not tested and later develops a serious disease? However, because EBM improves communication between doctors and patients and allows patients to share decision making, it protects doctors from litigation (most litigation happens when there is a Reference: breakdown in communication). EBM analyses have already been used in the courts and have been well accepted. Such empirical evidence has saved doctors from Del Mar C (2000). Asymptomatic trouble when opinion may have damned them. haematuria … in the doctor. British Medical Journal 320:165–166.16
  21. 21. Case study 4: painful shoulderA -year-old male complained of a painful left shoulder for several weeks. HisGP had often used cortisone injections for such shoulder pain, but was nownot sure if this was a good idea because she had seen a recent trial of cortisoneinjection for tennis elbow which showed good short-term improvementbut the long-term outcomes were worse than with watchful-waiting orphysiotherapy. A search of the Cochrane Library found a systematic reviewof randomised trials of several treatments for shoulder pain, which was lastupdated in .Based on two small trials (with a total of  patients), the authors concludedthat subacromial steroid injection showed some short-term benefit overplacebo. A further search of the Clinical Trials Registry identified a more recenttrial that compared physiotherapy, manipulation and corticosteroid injectionsin a total of  patients. It showed that corticosteroid injections had short-term benefits (up to  year) with a  absolute increase in ‘cure’ at  weeks.However, when long-term outcomes were measured (– years), about half thepatients had some recurrence and there was no difference between the threegroups. References:How did EBM help? Green S, Buchbinder R, Glazier R,The search revealed studies that answered the GP’s question and provided Forbes A (). Interventions foruseful information for the patient. The GP was able to advise her patient of shoulder pain (Cochrane Review).three things: In: The Cochrane Library, Issue , . Oxford: Update Software. Winters JC, Jorritsma W, Groenier KH• he would probably improve even without treatment et al (). Treatment of shoulder• a steroid injection would help to relieve pain in the short-term (up to  year) complaints in general practice: long term results of a randomised,• a steroid injection would make no difference in the long term (– years). single blind study comparing physiotherapy, manipulation, andBased on this information, the patient was able to make an informed decision corticosteroid injection. Britishabout whether to have the injection or not. Medical Journal :–. 17
  22. 22. Summary of case studiesThe case studies show that EBM has several advantages:• Medical practitioners, especially GPs, can’t know everything. EBM helps doctors keep up to date across a very wide spectrum of information.• MEDLINE and similar databases have several advantages. For medical practitioners, they are a way of finding up-to-date information that is not biased and is of good quality.• Because the search is based on questions rather than possible answers, doctors can find information without having heard about it before. In other words, they can find information that they do not initially know they need, but which, as we have seen, is important for good clinical practice.• The evidence can be used to quantify outcomes (empirical evidence). It allows people to assess the likelihood of benefiting from a particular treatment or activity rather than just considering the underlying mechanism.• Patients like this empirical approach because it is easier to understand and allows them to share in decision making.• Decision making can be shared between the doctor and patient based on empirical evidence of risks and benefits. This reduces the chances of future litigation.• Electronic searching can reveal other useful information that is of benefit to the patient.18
  23. 23. Participants’ own clinical questionsNow we will work out how to turn your day-to-day questions into structuredquestions that can be answered in a similar way to the case studies above.In the space provided below, write down a question in relation to eitheryourself or one of your patients. If you are stuck, write down the last patientyou saw and we will work out a question.In the next section (page 23), we will look at how to turn your questions into aform that can be used to search the medical literature in less than two minutes.Then we will use the computer lab to find answers to them. Write down a clinical question here 19
  24. 24. Notes20
  25. 25. EBM step 1: Formulate an answerable Steps in EBM:question 1. Formulate an answerable question. 2. Track down the bestFirst principle evidence of outcomesFirst, you must admit that you don’t know. available.As we have already seen, it is impossible to know everything. EBM gives you a 3. Critically appraise themethod to find answers to questions without having any prior knowledge of evidence (ie find out how good it is).what you ought to know. 4. Apply the evidence (integrate the results with clinical expertise and patient values). 5. Evaluate the effectiveness and efficiency of the process (to improve next time). 23
  26. 26. The ‘PICO’ principleQuestions often spring to mind in a form that makes finding answers in themedical literature a challenge. Dissecting the question into its component partsand restructuring it so that it is easy to find the answers is an essential first stepin EBM. Most questions can be divided into 4 parts: 1. The population or participants Who are the relevant patients? 2. The intervention or indicator What is the management strategy, diagnostic test or exposure that you are interested in (such as a drug, food, surgical procedure, diagnostic test or exposure to a chemical)? 3.The comparator or control What is the control or alternative management strategy, test or exposure that you will be comparing the one you are interested in with? 4. The outcome What are the patient-relevant consequences of the exposure in which we are interested?All clinical or research questions can be divided into these four components, whichwe call ‘P I C O’. It is important to use all four parts of the question, if possible.Remember the PICO principle P Population/patient I Intervention/indicator C Comparator/control O Outcome24
  27. 27. Different types of questionsBy far the most common type of clinical question is about how to treat adisease or condition. In EBM, treatments and therapies are called ‘interventions’and such questions are questions of INTERVENTION.However, not all research questions are about interventions. Other types ofquestions that may arise are as follows:1. What causes the problem? AETIOLOGY AND RISK FACTORS2. What is the frequency of the problem? FREQUENCY3. Does this person have the problem? DIAGNOSIS4. Who will get the problem? PROGNOSIS AND PREDICTIONIn each case the P I C O method can be used to formulate the question,as shown in the following examples. The same approach can be used toresearch qualitative questions about health issues of a more general nature(PHENOMENA). In this case, the question will consist of ‘P’ and ‘O’ only.The studies that you will need to search for are different for thedifferent types of questions and we will discuss this further in the next section(see ‘EBM step 2: Track down the best evidence’). 25
  28. 28. InterventionsInterventions cover a wide range of activities from drug treatments and otherclinical therapies, to lifestyle changes (for example, diet or exercise) and socialactivities (such as an education program). Interventions can include individualpatient care or population health activities (for example, screening for diseasessuch as cervical or prostate cancer).Example 1A 28-year-old male presents with recurrent furunculosis for past 8 months;these episodes have been treated with drainage and several courses ofantibiotics but keep recurring. He asks if recurrences can be prevented.To convert this to an answerable question, use the P I C O method as follows : P Population/patient = patients with recurrent furunculosis I Intervention/indicator = prophylactic antibiotics C Comparator/control = no treatment O Outcome = reduction in recurrence rate of furunculosisQuestion:‘In patients with recurrent furunculosis, do prophylactic antibiotics, comparedto no treatment, reduce the recurrence rate?’26
  29. 29. Example 2Jeff, a smoker of more than 30 years, has come to see you about somethingunrelated. You ask him if he is interested in stopping smoking. He tells you hehas tried to quit smoking unsuccessfully in the past. A friend of his, however,successfully quit with accupuncture. Should he try it? Other interventions youknow about are nicotine replacement therapy and antidepressants.Develop a clinical research question using P I C O: P Population/patient = I Intervention/indicator = C Comparator/control = O Outcome =Question: 27
  30. 30. Example 3At a routine immunisation visit, Lisa, the mother of a -month-old, tells youthat her baby suffered a nasty local reaction after her previous immunisation.Lisa is very concerned that the same thing may happen again this time.Recently, a colleague told you that needle length can affect local reactions toimmunisation in young children but can’t remember the precise details.Develop a clinical research question using P I C O to help you find theinformation you need: P Population/patient = I Intervention/indicator = C Comparator/control = O Outcome =Question:28
  31. 31. Example 4In browsing one of the medical weeklies, you come across mention ofimiquimod cream for treatment of basal cell carcinomas (BCC). The idea ofa cream for BCCs is surprising, so you wonder about the effectiveness andparticularly the long-term cure rate of imiquimod cream.Develop a clinical research question using P I C O to help answer your query: P Population/patient = I Intervention/indicator = C Comparator/control = O Outcome =Question: 29
  32. 32. Aetiology and risk factorsQuestions of aetiology and risk factors are about what causes a disease orhealth condition. They are the reverse of intervention questions because theydeal with the harmful outcomes of an activity or exposure. Such questionscommonly arise in relation to public health issues, such as whether eatingcertain foods increases the risk of heart disease, or being exposed to anenvironmental chemical increases the risk of cancer, and so on.Example 1George has come in to your surgery to discuss the possibility of getting avasectomy. He says he has heard something about vasectomy causing anincrease in testicular cancer later in life. You know that the risk of this is low butwant to give him a more precise answer. P Population/patient = adult males I Intervention/indicator = vasectomy C Comparator/control = no vasectomy O Outcome = testicular cancerQuestion:‘In men, does having a vasectomy (compared to not having one) increase therisk of getting testicular cancer in the future?’30
  33. 33. Example 2Susan is expecting her first baby in two months. She has been reading aboutthe potential benefits and harms of giving newborn babies vitamin K injections.She is alarmed by reports that vitamin K injections in newborn babies maycause childhood leukaemia. She asks you if this is true and, if so, what the riskfor her baby will be.Develop a clinical research question using P I C O to help answer Susan’squestion: P Population/patient = I Intervention/indicator = C Comparator/control = O Outcome =Question: 31
  34. 34. Frequency or rateQuestions of frequency (prevalence) are about how many people in thepopulation have a disease or health problem, such as what is the frequency ofhearing problems in infants or the prevalence of Alzheimer’s disease in the over70s. If the question also includes a time period, such as for cases of influenza inwinter versus summer, it becomes a question of rate (incidence).Example 1Mabel is a 6-week-old baby at her routine follow-up. She was born prematurelyat 35 weeks. You want to tell the parents about her chances of developinghearing problems. P Population/patient = infants I Intervention/indicator = premature C Comparator/control = full-term O Outcome = sensorial deafnessQuestion:‘In infants born prematurely, compared to those born at full term, what will theprevalence of sensorial deafness be?’32
  35. 35. Example 2Mrs Smith has acute lower back pain. She has never had such pain before andis convinced that it must be caused by something really serious. You take ahistory and examine her but find no indicators of a more serious condition. Youreassure her that the majority of acute low back pain is not serious but she isstill not convinced.Develop a clinical research question using P I C O to help reassure Mrs Smith: P Population/patient = I Intervention/indicator = C Comparator/control = O Outcome =Question: 33
  36. 36. DiagnosisDiagnosis questions are concerned with how accurate a diagnostic test is invarious patient groups and in comparison to other available tests. Measures oftest accuracy include its sensitivity and specificity.Example 1Julie is pregnant for the second time. She had her first baby when she was 33and had amniocentesis to find out if the baby had Down syndrome. The testwas negative but it was not a good experience as she did not get the result untilshe was 18 weeks pregnant. She is now 35, one month pregnant and asks if shecan have a test that would give her an earlier result. The local hospital offersserum biochemistry plus nuchal translucency ultrasound screening as a firsttrimester test for Down syndrome. You wonder if this combination of tests is asreliable as conventional amniocentesis. P Population/patient = pregnant women I Intervention/indicator = nuchal translucency ultrasound screening plus serum biochemistry (first trimester) C Comparator/control = conventional amniocentesis O Outcome = accurate diagnosis (measured by sensitivity and specificity) of Down syndrome (trisomy 21)Question:‘For pregnant women, is nuchal translucency ultrasound screening plus serumbiochemistry testing in the first trimester as accurate (ie with equal or bettersensitivity and specificity) as conventional amniocentesis for diagnosing Downsyndrome?’34
  37. 37. Example 2As part of your clinic’s assessment of elderly patients, there is a check ofhearing. Over a tea room discussion it turns out that some people simply askand others use a tuning fork, but you claim that a simple whispered voice testis very accurate. Challenged to back this up with evidence, you promise to do aliterature search before tomorrow’s meeting.Develop a clinical research question using P I C O to help you with yourliterature research: P Population/patient = I Intervention/indicator = C Comparator/control = O Outcome =Question: 35
  38. 38. Prediction (prognosis)Prediction or prognosis questions are concerned with how likely an outcome isfor a population with certain characteristics (risk factors), such as the likelihoodthat a man who is experiencing atypical chest pains will suffer further heartfailure or sudden death within the next few days, or the predicted morbidityand mortality for a person diagnosed with colon cancer.Example 1Childhood seizures are common and frightening for the parents and thedecision to initiate prophylactic treatment after a first fit is a difficult one.To help parents make their decision, you need to explain the risk of furtheroccurrences following a single seizure of unknown cause. P Population/patient = children I Intervention/indicator = one seizure of unknown cause C Comparator/control = no seizures O Outcome = further seizuresQuestion:‘In children who have had one seizure of unknown cause, compared withchildren who have had no seizures, what is the increased risk of furtherseizures?’36
  39. 39. Example 2Mr Thomas, who is  years old, has correctly diagnosed his inguinal lump as ahernia. He visits you for confirmation of his diagnosis and information aboutthe consequences. You mention the possibility of strangulation, and the manasks: ‘How likely is that?’ You reply ‘pretty unlikely’ (which is as much as youknow at the time) but say that you will try and find out more precisely.Develop a clinical research question using P I C O to help you give Mr Thomasmore precise details about his prognosis: P Population/patient = I Intervention/indicator = C Comparator/control = O Outcome =Question: 37
  40. 40. PhenomenaQuestions about phenomena can relate to any aspect of clinical practice,such as physical examination, taking a health history or barriers to successfulparticipation in health care. Such questions usually involve a population (P)and an outcome (O) but not an intervention or comparator.Example 1Mary is a mother who is concerned about her child of 3. He has a fever. Afteryou have examined him you conclude that he probably has a viral infection.Mary asks ‘But what if he has a fever again during the night doctor?’ You wantto understand her principle underlying concerns so that you can reassure her. P Population/patient = mothers of children with fever O Outcome = principle concernsQuestion:For mothers of children with a fever, what are the principle concerns?Example 2When giving immunisation injections to children, you notice that many of thechildren make distinctive facial expressions when they receive the injection andyou wonder how these expressions are related to their experience of pain. P Population/patient = O Outcome =Question:38
  41. 41. Your own questions1. Write here the clinical issue that you wrote down earlier (page 19).Identify what sort of question it is (circle):intervention phenomenon aetiology frequency diagnosis predictionNow build up a research question using P I C O P Population/patient = I Intervention/indicator = C Comparator/control = O Outcome =Question: 39
  42. 42. Your own questions2. Write a second clinical issue that interests you.Identify what sort of question it is (circle):intervention phenomenon aetiology frequency diagnosis predictionNow build up a research question using P I C O P Population/patient = I Intervention/indicator = C Comparator/control = O Outcome =Question:40
  43. 43. Types of studiesThe types of studies that give the best evidence are different for the differenttypes of questions. In every case, however, the best evidence comes fromstudies where the methods used maximise the chance of eliminating bias. Thestudy designs that best suit the different question types outlined are as follows: Question Best study designs Description INTERVENTION Randomised controlled trial Subjects are randomly allocated to treatment or control groups and outcomes assessed. AETIOLOGY AND Randomised controlled trial As aetiology questions are similar to intervention questions, the RISK FACTORS ideal study type is an RCT. However, it is usually not ethical or practical to conduct such a trial to assess harmful outcomes. Cohort study Outcomes are compared for matched groups with and without exposure or risk factor (prospective study). Case-control study Subjects with and without outcome of interest are compared for previous exposure or risk factor (retrospective study). FREQUENCY Cohort study As above AND RATE Cross-sectional study Measurement of condition in a representative (preferably random) sample of people. DIAGNOSIS Cross-sectional study with Preferably an independent, blind, comparison with ‘gold standard’ random or consecutive test. sample PROGNOSIS AND Cohort /survival study Long-term follow-up of a representative cohort. PREDICTION PHENOMENA Qualitative Narrative analysis, or focus group; designed to assess the range of issues (rather than their quantification).In each case, a systematic review of all the available studies is better than anindividual study. 41
  44. 44. Notes42
  45. 45. EBM step 2: Track down the best Steps in EBM:evidence 1. Formulate an answerable question. 2. Track down the bestWhere to search evidence of outcomesThe two main databases of information that we will use to search for evidence available.are: 3. Critically appraise the evidence (ie find out howPubMed good it is). 4. Apply the evidenceNational Library of Medicine free internet MEDLINE database. (integrate the results with clinical expertise andhttp://www.ncbi.nlm.nih.gov/entrez/query.fcgi patient values).The ‘Clinical Queries’ section of PubMed is a question-focused interface with 5. Evaluate the effectivenessfilters for identifying the more appropriate studies for questions of therapy, and efficiency of theprognosis, diagnosis and aetiology. process (to improve next time).The Cochrane LibraryThe Cochrane Library contains all the information collected by the CochraneCollaboration. It contains the following databases:The Cochrane Database of Cochrane systematic reviewsSystematic ReviewsThe Cochrane Controlled Trials Register of clinical trials that have been carriedRegister out or are in progress. The register contains over 300,000 controlled trials, which is the best single repository in the world.The Database of Abstracts Structured abstracts of systematic reviewsof Reviews of Effectiveness(DARE)Access to the Cochrane Library is free for users in many countries.http://www.cochrane.org and follow the promptsOther useful places to search are shown in the ‘Resources and further reading’section of this workbook. 43
  46. 46. The question guides the searchIn the previous section we discussed how to break down any type of clinicalquestion into four components: P Population/patient I Intervention/indicator C Comparator/control O OutcomeYou can now use these components to direct your search. It is also worthlooking for synonyms for each component.General structure of question (Population OR synonym1 OR synonym2…) AND (Intervention OR synonym1 OR synonym2…) AND (Comparator OR synonym1 OR synonym2…) AND (Outcome OR synonym1 OR synonym2…)Example: Question: In adults screened with faecal occult blood-testing, compared to no screening, is there a reduction in mortality from colorectal cancer? Question part Question term Synonyms Population/setting Adult, human – Intervention or indicator Screening, colorectal Screen, early detection, cancer bowel cancer Comparator No screening – Outcome Mortality Death*, survival* = wildcard symbol (finds words with the same stem)44
  47. 47. The parts of the question can also be represented as a Venn diagram: Mortality Screen Colorectal neoplasmOnce the study question has been broken down into its components, they canbe combined using the Boolean operators ‘AND’ and ‘OR’. For example: — represents the overlap between these two terms — retrieves only articles that use both terms. Remember: — represents the small area where all three circles overlap — retrieves only articles with all three terms.Complex combinations are possible. For example, the following combination ORcaptures all the overlap areas between the circles in the Venn diagram: Retrieves all articlesAlthough the overlap of all the parts of the question will generally have the with either wordbest concentration of relevant articles, the other areas may still containmany relevant articles. Hence, if the disease AND study factor combination(solid circles in Venn diagram) is manageable, it is best to work with this andnot further restrict by, for example, using outcomes (dotted circle in Venndiagram). ANDWhen the general structure of the question is developed it is worth looking forsynonyms for each component.Thus a search string might be: Retrieves only articles with both words 45
  48. 48. The term ‘screen*’ is shorthand for words beginning with screen, for example,screen, screened, screening. (Note: the ‘wildcard’ symbol varies betweensystems, eg it may be an asterisk [*], or colon [:].)In looking for synonyms you should consider both textwords and keywords inthe database. The MEDLINE keyword system, known as MeSH (Medical SubjectHeading), has a tree structure that covers a broad set of synonyms very quickly.The ‘explode’ (exp) feature of the tree structure allows you to capture an entiresubtree of MeSH terms within a single word. Thus for the colorectal cancerterm in the above search, the appropriate MeSH term might be:with the ‘explode’ incorporating all the MeSH tree below colonic neoplasm, viz:While the MeSH system is useful, it should supplement rather than replace theuse of textwords so that incompletely coded articles are not missed.The MeSH site can be accessed from PubMed (see ‘How to use PubMed’ laterin this section).46
  49. 49. Searching tips and tacticstruncation and wildcard (*) NEAR = AND plus words close togetherBOOLEANS IN CAPITALS Word must be in TITLE Group words with ( ) Finds studies containing either of the specified words or phrases. For example, finds articles with either the word child or the word adolescent. Finds studies containing both specified words or phrases. For example, finds articles with both the word child and the word adolescent. Like , requires both words but the specified words must also be within about 5 words from each other. Excludes studies containing the specified word or phrase. For example, means studies with the word ‘child’ but not the word ‘adolescent’. Use sparingly. Articles retrieved may be restricted in several ways, eg by date, by language, by whether there is an abstract, etc. Use parentheses to group words. For example, finds articles with one or both ‘child’ and ‘adolescent’ and one or both of the words ‘hearing’ or ‘auditory’. Truncation: the ‘ ’acts as a wildcard indicating any further letters, eg child is child plus any further letters and is equivalent to . Finds studies with the word in the title. For example, (in PubMed) and (in Cochrane) finds studies with the word hearing in the title. Retrieves studies from a specific source, eg finds articles on hearing in the BMJ. MeSH is the Medical Subject Headings, a controlled vocabulary of keywords which may be used in PubMed or Cochrane. It is often useful to use both MeSH heading and text words. 47
  50. 50. Computer searchingWe ‘bookmark’ our computers to rapidly go to the Cochrane Library andPubMed. Where you search first depends on the type of question you haveasked. For an intervention question, the best evidence comes from a systematicreview of RCTs. Therefore, first check the Cochrane Database of SystematicReviews within the Cochrane Library (see page 43). For other types of questionsyou should first search PubMed Clinical Queries.As a general rule you should start searching at the level that will give you thebest possible evidence. If you do not find anything, drop to the next level.For example, for an intervention question, first use the Cochrane Libraryto find out if there has been a systematic review of RCTs that relate to yourquestion. If there is a Cochrane systematic review on your question, this is thebest evidence that you will find anywhere, so you do not need to search otherdatabases because the evidence you find will not be as good as the Cochranesystematic review.If there is not a Cochrane systematic review, the Cochrane Library may still tellyou if there has been another quality systematic review (DARE database). ADARE review is the next best evidence after a Cochrane review so, if there isone you do not need to look further. If there is not a DARE review, check theCochrane Register of Controlled Trials (CENTRAL) to find out if there havebeen any RCTs (or if there are any in progress).If there is not a Cochrane systematic review, a DARE systematic review or evenan RCT in the Cochrane Library, you will need to go to PubMed to look forobservational data (such as case-controlled studies, cohort studies, or evencase series). Start by going to the ‘Clinical Queries’ section because this willbring up the most clinically relevant studies. The search path you can follow formost questions is shown in the following flowchart.48
  51. 51. Type of question Intervention Other Cochrane LibraryCochrane No Cochranesystematic systematic review review Database of Abstracts STOP of Reviews of Effectiveness (DARE) Systematic No systematic review review Central Register of Controlled STOP Trials (CENTRAL) RCTs No RCTs PubMed Clinical Queries CRITICAL APPRAISAL RCTs/cohort No studies studies etc. STOP PubMed general CRITICAL APPRAISAL search/other databases STOP BEWARE you are now entering territory with lower quality studies! 49
  52. 52. How to use the Cochrane LibraryGo to the Cochrane Library homepage at:http://www.cochrane.org and follow the prompts.If you are in a registered country for use of the library you can use the ‘Log onanonymously’ button to log into the library.The library includes:• The Cochrane Database of Systematic Reviews with approximately 1500 completed reviews and 1100 protocols (reviews that are currently in progress but not finished)• Database of Abstracts of Reviews of Effectiveness (DARE) with about 3000 abstracts of other systematic reviews• The Cochrane Central Register of Clinical Trials (CENTRAL), which lists over 350,000 controlled trials that have been carried out or are currently in progress, many with abstracts.The screenshot on this page is reproduced with permission from update software50
  53. 53. To search the library, enter your search phrase in the space provided. Theresults will show the total ‘hits’ from the site and the hits from each database.Click on each report to show the details.Cochrane systematic reviews are very detailed but each has a structuredabstract with the main findings. You can also go to the ‘Tables and Graphs’section towards the end of the report and click on the studies to see the resultsof the analysis. These results can often be used to calculate a ‘number neededto treat’ (NNT).For example, the search terms ‘ ’shows the following systematic review:‘Local corticosteroid treatment for carpal tunnel syndrome’The ‘Tables and Graphs’ area of the review shows one study wherecorticosteroid treatment was compared to placebo treatment as thecomparator with the numbers of patients showing improvement at 1 month asthe outcome.The results showed a statistically significant benefit at one month for thetreated group of patients as follows: Number improved % improved at 1 month Hydrocortisone 23/30 77 Placebo 6/30 20 Percentage improved 57 (57 better from 100 treated) because of treatment NNT 100/57 = 1.75 (ie more than 1 in every 2 patients treated will improve)(Note: Symptom improvement beyond 1 month has not been demonstrated) 51
  54. 54. How to use PubMedGo the the ‘Entrez-PubMed’ webpage at:http://www.ncbi.nlm.nih.gov/entrez/query.fcgiYou can search directly from the entry page by typing your search terms intothe box at the top. Click ‘Limits’ to set limits such as date, language and type ofarticle. However, this sort of search does not provide any filtering for quality ofthe research and you will probably retrieve a large number of articles of variableusefulness.To improve the quality of the studies you retrieve, click on ‘Clinical Queries’ onthe sidebar.The screenshots on p – are reproduced with permission. Source: The National Center forBiotechnology Information, The National Library of Medicine, The National Institute of Health,Department of Health and Human Services52
  55. 55. Next, enter the type of question you are trying to answer (ie intervention[therapy], diagnosis, aetiology, prognosis). If you click the ‘Sensitivity’ buttonyou will get more articles but some may be less relevant. ‘Specificity’ gives youonly highly relevant articles.Finally, enter your search terms in the box and click ‘Go’. 53